Create translation.py

pipeline/translation.py

@@ -0,0 +1,10 @@
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+
+_model_name = "Helsinki-NLP/opus-mt-en-hi"
+_tokenizer = AutoTokenizer.from_pretrained(_model_name)
+_model = AutoModelForSeq2SeqLM.from_pretrained(_model_name)
+
+def translate_text(text: str) -> str:
+    inputs = _tokenizer(text, return_tensors="pt", truncation=True)
+    outputs = _model.generate(**inputs, max_new_tokens=200)
+    return _tokenizer.decode(outputs[0], skip_special_tokens=True)